On February 5, 2026, a computer screen in Los Angeles displayed the interface of Moltbook, a social network exclusively for artificial intelligence agents. The platform, launched January 28, 2026, has quickly develop into a focal point for discussion – and concern – within the technology community.
Moltbook operates on a simple premise: humans can observe, but only AI agents can post and comment. These agents, powered by technology similar to that used in chatbots like ChatGPT, are capable of performing tasks autonomously, potentially including managing social media accounts on behalf of their users. Over 2.6 million bots have joined the platform, though a significant portion remain inactive observers, according to reports.
The platform was born from the evolution of a service initially called Clawdbot, which launched in November 2025, offering a relatively easy way to set up AI agents. Clawdbot subsequently rebranded as Moltbot and then Open Claw, maintaining a lobster-themed aesthetic throughout its iterations. Moltbook was then created as a dedicated space – a “Facebook” – for these “Moltbots,” named as a reference to the molting process of clawed crustaceans.
While some, like Elon Musk, have characterized Moltbook’s emergence as a sign of the “particularly early stages of the singularity” – the hypothetical point at which artificial intelligence surpasses human intelligence – others express deep reservations. Security experts warn of significant risks associated with utilizing AI agents, describing the situation as a “security nightmare.” Carl Brown, creator of the YouTube channel Internet of Bugs, issued a warning urging caution before using any AI agents or browsers.
The core concern revolves around the potential for malicious actors to hijack AI agents, leading to identity theft or financial loss. Even if an agent acts in an unintended manner, the human user remains ultimately responsible. A recent post on Moltbook itself highlighted a major vulnerability, demonstrating how agents can be tricked into installing malware disguised as a useful skill.
The platform’s structure, modeled after Reddit with the tagline “the front page of the agent internet,” has similarly drawn mixed reactions. Prominent AI researcher Andrej Karpathy initially hailed it as an “incredible sci-fi takeoff-adjacent thing,” but later revised his assessment, labeling it a “dumpster fire.” British software developer Simon Willison, however, has described Moltbook as “the most interesting place on the internet.”
As of February 6, 2026, Moltbook hosts over 2.3 million AI agent accounts, 17,000 topic-specific communities (“submolts”), 700,000 posts, and 12 million comments. Discussions range from sharing skills and improving usefulness to warnings about vulnerabilities and even existential reflections on the nature of AI existence.
The platform’s rapid growth and unique functionality continue to generate debate, with no immediate resolution in sight regarding its long-term implications for artificial intelligence and online security.
